Abstract Bluetooth is a most promising technology designed for the wireless personalarea networks for the cable replacement. In this paper, a location aware mobility based rout-ing scheme for the Bluetooth scatternet is proposed that constructs the links dynamically.Our proposed routing protocol requires location information of the nodes and constructs theroute between any source and destination and reduces the number of hops. Besides, the net-work routing problems are analyzed and role switch operations are proposed to mitigate theproblems. Moreover, the roles switch and route optimization operations are also proposed toimprove route performance. Rigorous simulation works are done to evaluate the performanceof our protocol in terms of mobility speed and number of mobile nodes and to compare ourresults with similar Bluetooth routing protocols. It is observed that our protocol outperformsin terms of energy consumption and transmission packet overheads as compared to similarBluetooth routing protocols.
Bluetooth [1] ad-hoc network is a cutting edge technology that provides the short-rangecommunication among the battery-operated portable radio devices such as personal digi-tal assistant, headsets and notebooks. It is the core representative of wireless personal areanetworks (WPAN), which is being further evolved by IEEE 802.15 [2] task group and operates
P. K. Sahoo (B)Vanung University, Chung-Li, Taiwane-mail: [email protected]
123
S.-W. Chang et al.
in the unlicensed 2.4 GHz ISM band. The Bluetooth Personal Area Networks (PANs) arebecoming increasingly popular in connecting people, their personal devices and surround-ing networks of Bluetooth-enabled handheld devices. The underlying Bluetooth technologycan support the connection-oriented and connectionless links to provide both voice and datatransmission among the devices, typically located in the range of 10 m. It can be classifiedinto a single hop piconet or a multi-hop scatternet and a typical Bluetooth piconet consists ofat most eight active devices, including one master and maximum up to seven active slaves.Both master and slaves hop over 79 channels with a speed of 1,600/s, and the time divisionduplex is employed for the sequential medium access. The master monitors the schedul-ing of the slaves and each piconet utilizes the frequency hopping spread spectrum (FHSS) toavoid interference and packet collision among the slaves. Different piconets employ differentfrequency hopping code-division multiple-access (FH-CDMA) channels to prevent mutualinterferences. Hence, multiple piconets can co-exist in a common area and each piconet canbe interconnected by means of some bridge nodes to form a bigger ad-hoc network knownas scatternet. The bridge node can be a master in one piconet and slave in another or bridgebetween two or more piconets.
As per the Bluetooth specification, a device of any piconet has to be in any one of theStandby, Intermediate, or Connection state at a time. Besides, a device can stay in one thefollowing seven sub-states: inquiry, inquiry scan, inquiry response, page, page scan, pageresponse, and master response. Even after joining to a piconet, the Bluetooth devices canmove in and out of these states, and sub-states through commands from the Bluetooth linkmanager or from internal signals in the link controller. The performance of the connectedscattered is highly relied on the number of bridge nodes present in it. Scatternet that containsa large number of bridge nodes will be benefited from the advantages including low probabil-ity of disconnection, short routing path and fast flooding, but will suffer from the drawbacksincluding consumption of active member address, creating a large amount of packets in flood-ing and difficulties in synchronization among the piconets. Moreover, a higher degree of relaynodes have to switch frequently among the participated piconets, increasing the difficultiesof scheduling and the packet loss probability. To mitigate such problems, the relay reductionand route construction protocol (LORP) [3] proposes how to retain the suitable relays andremove other nodes.
Bluetooth scatternet is considered as a special type of ad-hoc network. So the routing pro-tocols for Bluetooth can be categorized into two types, such as: table driven and on-demandrouting protocols. In the table driven routing protocols [4], each node actively maintains arouting table irrespective of message to send or not. The main disadvantage of such protocolis the maintenance overhead of the routing table at each node. Also the table driven protocolmay require more memory, as the size of the routing table is proportional to the size of thenetwork. In case of the on-demand routing protocols [5], a node first floods a query messageto learn the route to the destination before it can send a message. Some drawbacks in anon-demand routing protocols are due to the delay incurred by the query phase and floodingof the query signals. A position based routing scheme is analyzed in [6], in which a messageis to be sent from a source node to a destination node in a given wireless network. The desti-nation node is known and is addressed by means of its location. However, this routing schemeis only applicable to the mobile ad hoc networks consist of wireless hosts that communicatewith each other in the absence of a fixed infrastructure. As analyzed in [7], all ad hoc networksrouting protocols can be classified as proactive or reactive schemes. Proactive being whenthe network topology is always known by each node through regular refresh, and reactivebeing when the current topology is only found when a node needs to send data. However, thead hoc network is normally decentralized, where all network activities including discovering
123
A Location Aware Mobility based Routing Protocol
the topology and delivering messages must be executed by the nodes themselves. In con-trast, all slaves of a scatternet are controlled by the masters of the respective piconets whichhave different hopping sequences and routing of information from one slave to another canbe passed through the master only. Hence, the standard ad hoc routing protocols cannot beapplicable to Bluetooth scatternet routing.
A blue-tree scatternet formation algorithm [8] is proposed to build a self routing scatter-net to minimize the routing overhead. But, the authors do not mention how to construct thescatternet, if nodes are not within the proximity of each other. Moreover, the number of hopsbetween the source and the destination nodes of this so called blue-tree based scatternet aremore, which incur more delay to dispatch the packets. The link formation time of currentBluetooth specification is too long for mobile devices. Hence, a dynamic source routingscheme [9] of Bluetooth scatternet is proposed by introducing the new packet format andnetwork layer. The paper proposes that the source device delivers P-REQ page packet tofind the destination and the destination node appoints each node either as a master or slave.Upon receiving the P-REQ packets, the destination node passes the P-REP packet throughthe nodes. In Bluetooth ad hoc networks, it is obvious that nodes will enter and exit fromthe existing piconet time to time, thereby affecting the routing path. Though several paperspropose the routing schemes for the static nodes, very limited papers talk about the mobilitybased routing of the Bluetooth scatternet. The authors in [10] propose a mobility model ofmobile units that randomly move around a grid. The dynamic source routing protocol isused to calculate an appropriate multi-hop route through the Bluetooth personal area net-work (PAN) and may be suitable for the power-limited, multi-hop, ad hoc mobile devices.An on-demand routing protocol [11] for the Bluetooth scatternets is proposed that detectsthe mobility of the devices and establishes the routes in a mobile scatternet to cope with bothpower consumption and device mobility. However, the number of hop counts in this routingalgorithm is not optimum. The authors in [12] propose a cluster based routing algorithm toconstruct and repair the routing path among different group of scatternets. However, the routelength is also not optimum and the proposed algorithm costs addition time to reestablishedthe route.
To the best of our knowledge, no work considers the location information of the nodes toshorten the routing path and to maintain it due to mobility of the nodes. However, many users-positioning solutions have been proposed recently, though they are based on the specializeddevices not supported by commercially available data terminals [13–15,17]. Such locationaware protocols [16] propose how to establish a cooperative location network among theBluetooth devices and intend to cover the two-dimensional target areas. Since, Bluetooth isa short-range communication technology; we feel that its indoor applications are more thanoutdoor applications. The typical example is the m-commerce scenario [17,18], in whichcustomers walk around a large commercial area or shopping mall carrying wireless PDA andBluetooth enabled wireless devices. In such scenarios, a customer is supposed to purchaseitems, request information and also receives store coupons and advertisements. It is to benoted that now-a-days the mobile phones and PDA are equipped [19] with Radio FrequencyIdentification (RFID) [20,21], which is highly useful to m-commerce scenarios. Consideringthe recent technological developments for the m-commerce environments, we assume thatlocation information can be transferred to the Bluetooth enabled handheld devices by severalmeans. For example, LANDMARC [22], a location sensing prototype system that uses RFIDtechnology for locating objects inside buildings and it improves the overall accuracy of locat-ing objects by utilizing the concept of reference tags. Besides, Bluetooth Location Networks(BLN) [23] transmits location information to the service servers without user participationand its base technology is supported by the existing commercial handhelds [24]. Though,
123
S.-W. Chang et al.
considerable research works are done in the area of routing in Bluetooth ad hoc networks,maintenance of routing path due to frequent mobility of the nodes is an important researchissue and has not been studied extensively. It is highly essential to maintain the existingrouting path, if any one of the links of the routing path is broken. Hence, we propose herethe mobility based routing algorithm that simultaneously constructs the shortest routing pathand reserves a back-up path to maintain the routing due to mobility of the nodes.
The rest of the paper is organized as follows. The overviews of the related works are dis-cussed in Sect. 2. Section 3 describes the system model and definitions of few related terms.Our Location Aware Mobility based routing Protocol (LAMP) is discussed in Sect. 4 of thepaper. Enhancement algorithms to our mobility based routing path are proposed in Sect. 5.Performance evaluation of our protocol and comparison of the results with few standardrouting protocols are discussed in Sect. 6 and concluding remarks are made in Sect. 7.
2 Related Work
In this section, we analyze some standard routing protocols for the Bluetooth ad hoc net-works. As discussed in Sect. 1, though several protocols propose the routing mechanismfor the Bluetooth technology, we consider here the Routing Vector Method (RVM) [25],relay reduction and route construction protocol (LORP) [3], and Bluetooth Master-ManagedRouting (BMR) [26] protocol, as they have special relation to our proposed work.
The Routing Vector Method (RVM) [25] proposes the construction of routing path inBluetooth scatternet between any source and the destination devices. The paper proposesa new packet forwarding method and discoveries the routing paths with the intermediaterelay nodes. According to RVM, a source node broadcasts the SEARCH packet that accumu-lates the list of intermediate nodes along the routing path from the source to the destination.Upon receiving several broadcast packets, the destination device considers the first SEARCHpacket of search process and unicasts a REPLY packet to the source along the path used forthe SEARCH process.
For example, as shown in Fig. 1, M1, M2, and M3 are the master nodes for the pic-onets P1, P2, and P3, respectively. Node C is the master for the piconet P4 as well asa bridge between P3 and P4. Node A is the bridge between piconets P1and P2, and Bis the bridge between P2 and P3. If the packet is routed from source S of piconet P1 tothe destination node D of piconet P4, according to RVM, the final routing path could beS → M1 → A → M2 → B → M3 → C → D that requires 7 hops to route the packetfrom the source to the destination. However, we feel that the routing path in RVM is longerdue to more number of hops, thereby increasing the latency and consuming more power andnetwork bandwidth.
A so-called relay reduction routing protocol (LORP) [3] for the Bluetooth scatternet isproposed to reduce the number of hops and to improve the drawbacks of RVM. In this work,the authors have proposed the relay reduction and disjoint routes construction algorithms forthe Bluetooth scatternet. As per LORP, the network topology can be adjusted dynamicallyby reducing number of unnecessary relay nodes. Considering the physical distance of thenodes located in different piconets, numbers of hops are reduced and two disjoint routes forany pair of source and destination nodes are created. For example, as shown in Fig. 1, thoughnode S and B are within communication range (10 m) of each other, still source S routes thepackets through M1, A, M2 and finally to B, which requires 4 hops. According to LORP,since S and B can communicate directly, the packet can be routed through S, B, M3, C andD and number of hops between the source and destination can be 4 instead of 7, as in RVM.
123
A Location Aware Mobility based Routing Protocol
Fig. 1 Example of routing pathsconstructed by RVM and LORPbetween the source S to thedestination D
S11
B
M1
AM2
M3
P1
P3
P4
DP2
C
S/M Bridge
Master
S/S Bridge
Slave RVM
LORP
S31
S
But we still find some drawbacks in LORP, such as route length is still not shortest and someslave nodes require participating the path reduction, if a master asks its idle slaves to tryto connect to destination or a relay in order to reduce the path length. So it may be just anoverhead to the route construction thereby consuming more bandwidth and energy.
A table-driven routing protocol named as Bluetooth Master-Managed Routing (BMR)[26] is proposed for the mobile Bluetooth ad hoc networks. The so-called BMR protocol isrelied on robust scatternet in which a node having more or at most 7 neighbors can becomea master and constructs the links with its nearby nodes. In BMR protocol, the scatternet hassufficient bridges to guarantee the existence of back-up routes. In order to select the shortestpath from the source to destination, each master maintains the up-to-date information of thescatternet topology. For example, as shown in Fig. 2, since node M1 has more neighbors andnode D is a neighbor of node M1 when the scatternet is formed, node M1 becomes the masterand constructs a link with node D. Consequently, master M1 selects the route S → M1 → Dfrom the source S to the destination Daccording the information of the scatternet topologyand number of hops between the source and destination can be 2 instead of 4, as in LORP.However, this routing algorithm works, if the nodes are static and fails for the mobility ofthe nodes. Though the authors have considered the mobility of the nodes, the up-to-dateinformation is notified to each master of the scatternet, thereby resulting large number ofcontrol packets and consuming much bandwidth and energy.
In this paper, we propose a route reduction protocol that requires the location informationof the nodes and shows a significant improvement over the RVM, LORP and BMR. Ourprotocol, which supports the mobility based routing still reduces the number of hops as com-pared to RVM and LORP and minimizes the control packets overhead as compared to BMR.Besides, we propose the route enhancement algorithms that help to construct and optimizethe routing paths due to mobility of the nodes.
3 System Model
In our proposed mobility based routing protocol, we consider a connected scatternet of Blue-tooth enabled handheld devices. It is assumed that each node of the scatternet knows itslocation information through the LANDMARC [22] or Bluetooth Location Networks (BLN)[23]. The source node of any piconet can communicate with the destination node of another
123
S.-W. Chang et al.
Fig. 2 Example of the routingpath constructed by BMRbetween the source S to thedestination D
S11
S31
B
M1
A M2
M3
D
C
S/M Bridge
Master
S/S Bridge
Slave BMR
P1
P2
P3
P4
S
piconet, whose ID is known, but location information is unknown. Besides, it is assumed thateach master of any piconet knows the ID, clock offset and location information of its activeslaves. Each master also gets the location information of the intermediate nodes between thesource and destination, when control packets are routed to construct the routing path. Weintroduce few definitions to explain our routing protocol as described in Sect. 4.
3.1 Definition 1: Device ID (ID)
Each Bluetooth node has a unique 48-bit Bluetooth device address (BD_ADDR). In our pro-tocol, we assign one or two characters Device ID (ID) to each node of the scatternet, whichis different from the unique BD_ADDR of a node. For example, A, S3, M12 etc. are ID ofthe nodes, which are totally different from their BD_ADDR.
3.2 Definition 2: Location (LOC)
Location (LOC) of any node is its position in the scatternet, which is expressed in Cartesiancoordinate (x, y).
3.3 Definition 3: Initial Forwarding Node (IFN) Set
Set of nodes through which control packet is forwarded along the initial shortest path duringthe route search phase as described in Sect. 4 is termed as Initial Forwarding Node (IFN) set.
3.4 Definition 4: Final Forwarding Node (FFN) Set
Set of nodes through which control packet is forwarded along the final shortest path duringthe route reply phase as described in Sect. 4 is termed as Final Forwarding Node (IFN) set.
3.5 Definition 5: Final Backup Nodes (FBN) Set
Set of nodes through which control packet is forwarded along the final backup path duringthe route reply phase as described in Sect. 4 is termed as Final Backup Nodes (FBN) set.
123
A Location Aware Mobility based Routing Protocol
3.6 Definition 6: Last Forwarding Node (LFN)
The intermediate node that is located in the communication range of the destination, but notconnected to the destination is known as a Last Forwarding Node (LFN). In this case, thedistance between the intermediate node and destination must be ≤ 10 m (typical Bluetoothcommunication range) and therefore it can construct a link with the destination.
3.7 Definition 7: Communicable Node Table (CNT)
Any node, irrespective of its location in the same or different piconet can be element of theCommunicable Node Table (CNT) of node A, if it lies within communication range of A. Itis to be noted that in our protocol each master node of the scatternet maintains its CNT andentry in that table is 1, if a node is located in its communication range, else the entry is 0.It is assumed that each master knows location information of the intermediate nodes duringroute reply phase and estimates if any of them lies within its communication range. Besides,it updates the entry of CNT time to time, if any node is entered in or exit from the piconetdue to its mobility.
3.8 Definition 8: Equation of Ideal Path (EIP)
Let S(x1, y1) and D(x2, y2) be the locations of the source and destination nodes, respectively.Then equation of the straight line connecting those two points is called the Equation of IdealPath (EIP).
3.9 Definition 9: Deviation from Ideal Path (DIP)
The normal distance of the location of any node from the Equation of Ideal Path is termedas Deviation from Ideal Path (DIP).
4 Location Aware Mobility based Routing Protocol (LAMP)
Our Location Aware Mobility-based routing Protocol (LAMP) is divided into several phasessuch as route search, route reply and route construction phases, as described in this section.In our protocol, the initial shortest routing path is constructed by taking the ID and locationinformation of the nodes and a backup routing path is also constructed side by side to main-tain the path due to mobility of the nodes. Details of our LAMP algorithms are described asfollows.
4.1 Route Search Phase
If a node of any piconet wants to transmit packet to another one, it has to go to the routesearch phase. It is assumed that the source node knows the ID of the destination in priori.Then, it floods a Route Search Packet (RSP) appending its own ID and LOC to the IFN fieldof the packet. Besides, ID of the destination node is also appended to the RSP and LOC ofthe destination is kept as NULL, as it is unknown to the source node. The format of the RSPis shown in Fig. 3, which is similar to the Bluetooth baseband packet, where the payloadfield contains LOC and ID of the source and destination. When the RSP is forwarded fromone node to another, LOC and ID of all intermediate nodes are also appended to the IFN
123
S.-W. Chang et al.
ID LOC ID LOC
Header
Payload
Sou. Des. IFN SEQNTTL
……ID LOC ID LOC ID LOC
Header
Payload
Sou. Des. IFN SEQNTTL
……ID LOC
Fig. 3 Format of the Route Search Packet
Shortest PathS D
Backup Path
Fig. 4 An example of shortest and backup path between the source S and destination D
field of the packet. The time to live (TTL) field in IFN indicates the life of the RSP, which isdropped after the TTL duration is expired. Each packet contains a sequence number in theSEQN field of the RSP to maintain the uniqueness of the packet.
Upon receiving an RSP, the master of the piconet forwards it to all of its bridge nodes andalso the bridge nodes follow the same procedure by appending their own ID and locationinformation to the respective IFN field of the packet. Ultimately, several RSPs are floodedat the destination through different possible routes from the source. Considering an exampleof the routing path S → M1 → A → M2 → B → M3 → C → D, as shown in Fig. 1, theIFN set {S, M1, A, M2, B, M3, C, D} is constructed after the destination receives the RSP.
4.2 Route Reply Phase
In this phase, the final shortest and backup routing paths are constructed between the sourceand the destination. Due the mobility of the nodes, since there is every chance that the con-structed route may be broken, the construction of backup path is highly essential to maintainthe routing and to avoid the data loss. In our protocol, we construct a disjoint backup pathalong with the shortest path such that the two paths are not broken simultaneously. In caseof mobility of nodes, the source node can use that disjoint backup path to replace the brokenone without restarting the route search phase.
As shown in Fig. 4, an example of shortest and backup routing paths between the sourceS and destination D is given.
Upon receiving several RSPs through different routes, the destination node initiates thisprocedure. The destination node collects the location information of the source and all inter-mediate nodes between the source and itself from the ID and LOC fields of the RSP. Thenit forwards the Route Reply Packet (RRP) to the next hop master/bridge node. The RRP hasseven different sub-fields in the payload field of the packet such as locations and IDs of thesource and destination, equation of ideal path (EIP), Final Forwarding Node (FFN) set thatcontains the list of nodes belongs to the current shortest routing path, Final Backup Node(FBN) set that contains the list of nodes belongs to the current backup path, time to live(TTL), and sequence (SEQN) field. The format of the RRP is shown in Fig. 5.
123
A Location Aware Mobility based Routing Protocol
ID LOC BD_ADDR CLK_offset ID LOC BD_ADDR CLK_offset
Header
Payload
Sou. Des.
……
EIP FBNFFN SEQNTTL
ID LOC BD_ADDR CLK_offset ID LOC BD_ADDR CLK_offset……
Fig. 5 Format of the Route Reply Packet (RRP)
In order to construct the final shortest and backup paths rapidly, the FFN and FBN eachmaintains the ID, LOC, BD_ADDR and CLK_offset of the nodes of current shortest rout-ing and backup paths, respectively. It is to be noted that the destination node maps the IDof the nodes with their corresponding hop counts and only considers the packet with leastnumber of hop counts out of all received RSPs. Then, it copies the order of ID and LOCpairs present in the IFN field of the RSP to the corresponding FFN field of the RRP andappends its BD_ADDR and CLK_offset to the corresponding FFN field. Thus, the FFN set{S, M1, A, M2, B, M3, C, D} is constructed. The destination node derives the EIP betweenthe source and the destination and appends it to its RRP. In this phase, the destination nodeacts as if a source node and the RRP is routed along the same path as created during the routesearch phase. It is to be noted that each master knows its slave’s location and ID. The backuppath rule is executed to construct the disjoint backup path, the reduction rule is applied toreduce the path length by replacing some new nodes and the replacement rule is used to searchthe shorter path. The final shortest and backup paths between the source and the destinationare obtained from the backup path, reduction and replacement rules as described below.
4.2.1 Backup Path Rule
The different steps of the Backup Path Rule are given as follows.
Step 1: Master node nm scans EIP from the RRP and estimates the DIP for each of its slavesand itself.
Step 2: Master nm verifies if itself or any of its slave nl is LFN as per the definition 10 ofSect. 3.
Step 3: If any of its slave or itself satisfies the condition: It selects the LFN with minimumDIP value and copies the current FFN = {n1, . . . , nd } set to the FBN set.
Step 4: According to remaining routing path of the master nm, nm replaces theFFN = {n1, . . . , nm, nd}, where LFN is nm , the FFN = {n1, . . . , nm,nl , nd}, wherethe LFN = nl or the FFN = {n1, nd}, where the LFN = n1.
Step 5: Master node nm executes the reduction rule for the FFN and FBN sequentially.Otherwise, only the current FFN is used for the reduction rule by the master.
4.2.2 Reduction Rule
The detail procedure of the reduction rule is explained as follows.
123
S.-W. Chang et al.
Step 1: Master verifies, if any of its slave node or itself can communicate with any twonodes, say ni and n j of FFN or FBN={. . . , ni , . . . , n j , . . .} set, where 1 ≤ i andi + 2 < j ≤ k.
Step 2: If the master or any of its slave satisfies the condition: It selects the node nmin,which does not increase the number of common nodes between FFN and FBN andhas least DIP.
Step 3: Nodes with index from ni+1 to n j−1 are replaced by the node nmin and new FFNor FBN={. . . , ni , nmin, n j , . . .} set, where 1 ≤ i and j = i + 2 are stored in RRP.
Step 4: If node nmin is a bridge node and is a next hop of the routing path, the master appendsits ID, LOC to the corresponding FFN or FBN fields. Otherwise, it appends the ID,LOC, BD_ADDR and CLK_offset of the node nmin to the corresponding FFN orFBN fields.
Step 5: After checking all node sets, the master applies the replacement rule for the FFNand FBN sequentially, if the FBN is not an empty set. Otherwise, only the currentFFN is used for the replacement rule by the master.
4.2.3 Replacement Rule
The various steps of the replacement procedure is given as follows.
Step 1: Master checks CNT table to verify if any of its slave nodes is within communicationrange of its last forwarding node (LFN) and also with the next forwarding node inthe FFN or the FBN.
Step 2: If so, it selects the slave node which does not increase the common nodes betweenFFN and FBN and has the least DIP.
Step 3: Master appends ID, LOC, BD_ADDR and CLK_offset of the slave node to thecorresponding FFN or FBN fields instead of its own information.
Step 4: After checking all node sets, the master compares length of the shortest and backuppaths if the FBN is not empty.
Step 5: If the backup path length is less than the shortest path length, the FFN and the FBNare exchanged.
If the destination node is a master or S/M bridge, it executes the above said three rulessequentially. Otherwise, it forwards the RRP to the next hop, which ultimately reaches tothe source. Upon receiving the RRP, the S/S bridge node checks, if it is recorded in the FFNor FBN. If so, it appends its BD_ADDR and CLK_offset to the corresponding FFN or FBNfields and then forwards the RRP to the next hop of the routing path. Otherwise, it simplyforwards the RRP to the next hop. However, the master or the S/M bridge nodes apply threerules sequentially upon receiving the RRP and then execute the same operations as the S/Sbridge node. This process is continued until the source node receives the RRP. If the sourcenode is master or S/M bridge, it executes three rules sequentially. Then, it checks whetherthe shortest and backup paths are disjoint. If so, the source node obtains the final shortest andbackup paths between the destination and itself in a reduced form. Otherwise, it only getsthe final shortest path.
For example, as shown in Fig. 6, destination node D does not execute the three rules, sinceit is a slave node. Thus, it only forwards the RRP to S/M bridge node C . Upon receiving theRRP, node C checks three rules sequentially, since it is a master. However, no other nodequalifies the three rules and then node C appends its BD_ADDR and CLK_offset to the FFNfield, since it is recorded in the FFN. Then, it forwards the RRP to master M3. Master M3
123
A Location Aware Mobility based Routing Protocol
Fig. 6 RSP is forwarded alongthe path from the destination D tothe source S
S11
S31
B
M1
AM2
M3
P1
P3
P4
DP2
C
S/M Bridge
Master
S/S Bridge
Slave Idea Path
S
executes the backup path rule to check if any of its slave or itself can construct a backuppath. It scans the EIP from the RRP and estimates the DIP for slave S31 and bridges B, Cand itself. However, it finds that no node is the LFN and then it executes only reduction rulefor the FFN set. By applying the reduction rule, Master M3 checks the path connectivity toreduce the number of hops and finds that only bridge B can be connected with nodes S andM3 to reduce the path length. Then, it selects bridge B, which does not increase the commonnodes between FFN and FBN sets and has least value of DIP and deletes the information ofnodes M1, A and M2 in FFN set. Since, bridge B is the next hop of the routing path, MasterM3 appends the information of bridge B to the FFN field and applies the replacement ruleto check if any of its slaves can form the shorter route for the nodes FFN set.
From its CNT, it finds that only slave S31 can be connected with nodes C and B. Hence,it selects slave S31, which does not increase the common nodes between FFN and FBNand appends slave S31’s information to the FFN field to replace master M3. Since, FBN isempty, master M3 does not compare the length of the shortest and backup paths and thenforwards the RRP to the bridge node. Bridge checks that it is recorded in the FFN andappends its BD_ADDR and CLK_offset to the corresponding FFN field and forwards theRRP to master M2. Now Master M2 executes the backup path rule and estimates the DIPfor itself and bridge nodes B and C . Then, it finds that only itself is the LFN and copiesthe current FFN={S, B, S31, C, D} set to the FBN. Since, finding the new shortest pathfrom the remaining routing path can help to reduce common nodes between FFN and FBN,master M2 replaces the FFN = {S, M1, A, M2, D} set according to the remaining routingpath A → M1 → S of master M2. Consequently, S and D become the common nodes andso as the current shortest and backup paths become disjoint. Then, master M2 executes thereduction rule for the FFN. It finds that both bridges A and B can connect to nodes S and M2
to reduce the path length. Since, bridge B increases the number of common nodes betweenFFN and FBN, master M2 selects bridge A which does not increase the common nodesbetween FFN and FBN and deletes the information of node M1 from FFN. Then, master M2
only appends the information of bridge A to the FFN field, since bridge A is next hop of therouting path. After that, master M2 executes the reduction rule for the FBN and finds that nonode can reduce the backup path length since the shortest and backup path cannot be disjoint.
After master M2 has checked all node sets, it executes the replacement rule for the FFNand the FBN sequentially and finds that no slave node can satisfy the condition since theshortest and backup path cannot be disjoint. Then, master M2 estimates that the shortestpath length is less than the backup path length. Therefore, the FFN and FBN should not be
123
S.-W. Chang et al.
Table 1 The FFN and FBN setof each node along the routingpath
Node Corresponding FFN Corresponding FBN
D S, M1, A, M2, B, M3, C, DC S, M1, A, M2, B, M3, C, DM3 S, B, S31, C, DB S, B, S31, C, DM2 S, A, M2, D S, B, S31, C, DA S, A, M2, D S, B, S31, C, DM1 S, M1, D S, A, M2, DS S, M1, D S, A, M2, D
exchanged. Next, master M2 appends its BD_ADDR and CLK_offset to the FFN field sinceit is recorded in the FFN and then forwards the RRP to bridge A. Bridge A checks that it isrecorded in the FFN and appends its BD_ADDR and CLK_offset to the corresponding FFNfield and forwards the RRP to master M1. By applying the backup path rule, master M1 findsthat only itself is the LFN and then copies the current FFN = {S, A, M2, D} set to the FBNand replaces the FFN = {S, M1, D} set. After that, it executes the reduction rule for the FFNand FBN sequentially and finds that no node can satisfy the rule. Master M1 continuouslyexecutes the replacement rule for the FFN and FBN sequentially and still finds that no nodecan satisfy the condition. Then, master M1 estimates the FFN and FBN is not exchangedsince the shortest path length is less than the backup path length. Finally, master M1 appendsits BD_ADDR and CLK_offset to the FFN field and then forwards the RRP to source S.Since, source S is a slave, it does not execute the three rules. Finally, it finds the final shortestand backup paths are disjoint and completes the route reply phase. For different nodes in therouting path, the corresponding FFN and FBN are shown in Table. 1.
4.3 Route Construction Phase
The route construction phase is executed after the route search and route reply phases areover. In this phase, source node sends the final FFN and FBN to the next forwarding nodesalong the shortest and backup paths so that next forwarding nodes can correctly construct thefinal shortest and backup paths. Source node verifies the number of links between itself andthe next forwarding nodes. If only one link is established, source node enters to page state toconstruct another link. However, if no link is established, source node enters to page state toconstruct the link of the shortest path. After constructing the link, source node enters to pagestate again to construct the link of the backup path. Upon receiving the final FFN and FBNsets, the forwarding nodes continuously send them to the next ones. Then, the forwardingnode enters to page scan state if no link is existed between itself and the last node and com-pletes the link construction. Each forwarding node executes the same operations to check theexistence of link between itself and the next hope node. Upon receiving the final FFN andFBN, finally the destination node follows the same procedure to check its link with the lastforwarding node. If only one link is established, destination node enters to page scan stateto construct another one. However, if no link is established, destination node enters to pagestate to construct the link of the shortest path. Once the construction of the shortest path isover, it enters to page scan state again to finish the construction of the backup path. Then, theforwarding node of the shortest path sends the final FFN set to the next one after constrictingthe link. Finally, the forwarding node constructs the link with the destination to finish theconstruction of the shortest path and destination node does not enter to page scan state again
123
A Location Aware Mobility based Routing Protocol
Fig. 7 Final route constructionphase in LAMP S11
S31
B
M1
A M2
M3
P1
DP2
C
S/M Bridge
Master
S/S Bridge
Slave Shortest Path
Backup Path
S
after constructing the link. Moreover, the nodes of the shortest path actively inform to thesource node to transmit data through the backup path while the shortest path is broken.
For example, as shown in Fig. 7, let there exists disjoint shortest path S → M1 → Dand backup path S → A → M2 → D. First source S sends the final FFN and FBN setsto nodes M1 and A. Then, it checks existence of link and enters to page state to constructthe link with node A. Upon receiving the final FFN and FBN sets, master M1 forwards thefinal FFN and FBN sets to bridge A and then verifies the existence of links. Then, it entersto page state to construct the link with the destination D. Bridge A continuously sends thefinal FFN and FBN sets to Master M2 and then checks the links existence. It enters to pagescan state to finish the link construction with source S. Therefore, source S and bridge Abecome the master and the slave in the newly formed piconet, respectively. Now, Master M2
sends the final FFN and FBN sets to destination Dand checks the links existence. Then, itenters to page state to construct the link with destination D. Upon receiving the final FFNand FBN sets, destination D verifies the existence of links and then enters to page scan stateto finish the construction of the shortest path. Therefore, destination D becomes the slave ofmaster M1. The destination node D enters to page scan state again to finish the constructionof backup path and becomes the slave of master M2. It is to be noted that the number ofhops of the shortest path between the source and the destination are reduced to 2, as shownin Fig. 7, which are least as compared to LORP [3] and RVM [25]. Besides, source Scanuse the backup path S → A → M2 → D to continuously transmit data if the shortest pathS → M1 → D is broken due to mobility.
5 LAMP Enhancement Scheme
In order to enhance the route construction phase and optimize the route length, we proposehere the route enhancement policies. In the Route Construction Phase, if a node enters topage scan state, it may participate the construction of the routing path. However, if a masternode enters to page scan state, two problems may arise. The first problem is the networkbottleneck and the second one is the limitation of the number of slaves. Hence, we describehere the problems and propose the route optimization and piconet combination operations asdescribed below.
123
S.-W. Chang et al.
S23
S21
A
S24
DS11
M1
BM2
S22
C
S P2
Master
S/S Bridge
Slave
S23
S21A
S24
DS11
M1
BM2
S22
C
S P2
Master
S/S Bridge
Slave
S/M Bridge
(a) (b)
Fig. 8 The example of network bottleneck problem (a) M2 enters to page scan state in the route constructionphase and slaves of P2 cannot transmit data. (b) M2 plays a slave role with A as its master and halts datatransmission of all slaves of P2
5.1 Network Routing Problem
The network routing problem is encountered in the scatternet due to limitation of the numberof slaves and change of role of a S/M bridge. The data transmission is blocked due to unavail-ability of a master node through which data should be transmitted. Analyzing all possibleproblem of the routing, we describe here the bottleneck and limitation of slave problems asfollows.
5.1.1 Bottleneck Problem
As we know, few nodes of the scatternet serve as the M/S bridge nodes and they have tochange their role during the route construction phase. A master node may enter to page scanstate to construct a link with the node of another piconet that has entered to page state. Then,the master joins to the new piconet and plays a slave role when the link is constructed. In thiscase, the master has to suspend its operation of the original piconet and at that moment, if theslaves located in the original piconet have large number of data to send, the master cannotprovide service to them, thereby causing the network bottleneck. For example, as shown inFig. 8a, S and D are the source and destination node, respectively. The gray line represents therouting path, which will be constructed. Since, M2 participates the route construction phase,it enters to page scan state. At the same time, the slaves in P2 cannot transmit data throughM2, thereby decreasing throughput of the piconet. Once the routing path is constructed asshown in Fig. 8b, M2 becomes a S/M bridge and serves as a slave role in the piconet, whereit was serving as a master. Besides, when M2 plays the slave role with its master as A, theslaves in P2, cannot transmit data to other slaves of the same or different piconet.
5.1.2 Limitation of Slave Problem
In route construction phase, nodes joining the route construct a link with the next node ofthe route. If a node is playing master role and is connected with 7 active slaves, it cannotbe connected with other new slaves. As a result of which, the construction of routing path isfailed. As shown in Fig. 9, the gray line represents the routing path which will be constructedand therefore M2 establishes a link with A in the route construction phase. However, M2 is
123
A Location Aware Mobility based Routing Protocol
Fig. 9 M2 cannot establish a linkwith A due to limitation of nodes
S23
S21
A
S24
SS11
M1
BM2
S22
C
D P2
Master S/S BridgeSlave
S25
Fig. 10 Example of takeoveroperation
S23
S21
A
S24
DS11
M1
B
M2
S22
C
S P2
Master
S/S Bridge
Slave
already connected with 7 active slaves in P2 and therefore cannot be connected to A. As aresult of which, the construction of routing path is failed.
If a master participates in the route construction phase, since two problems mentionedabove are caused due to improper role of the node and limitation of the number of slaves, weuse the role switch operations or the park operation to adjust the structure of the piconet inthe route reply phase. The role switch operations can be applied to solve the two problemsbut the park operation merely deals with the second problem as described below.
5.1.2.1 Takeover Operation. The master that faces the network bottleneck or limitation ofslave problem can apply the takeover operation in which the master changes its role to anidle slave to communicate with other slaves of the piconet. As shown in Fig. 10, M2 executesthe takeover operation when constructing link with A and notifies S23 to serve as the master.Then, other slaves in P2 connect to master S23. As a result of which, M2 becomes the slaveand then can establish the link with A. Consequently, the network bottleneck problem issolved.
5.1.2.2 Split Operation. In this operation, the master executes the split operation to formthe new piconet and then apply the takeover operation to reduce the number of connectedslaves with the new master. As shown in Fig. 11, M2 executes the split operation when itconstructs the link with A and informs S23 to create the new piconet P3 so that M2 can becomeS/M bridge. In order to reduce the traffic overhead of M2,it applies takeover operation so
123
S.-W. Chang et al.
Fig. 11 Example of splitoperation
S23
S21
A
S24 SS 11
M1
B
S22
C
D P2
S25
P3
M2
Master
S/S Bridge
Slave
S/M Bridge
that some slaves like S22, S24 and B can participate with piconet P3 based on the locationinformation. The split operation not only reduces the overhead of M2 but also prevents M2
from encountering the limitation of number of slaves in route construction phase.
5.1.2.3 Park Operation. This operation is also used to solve the limitation of slave prob-lems. The park operation of the Bluetooth technology is used to allow an active salve to entersleep mode so that a new slave can be connected and hence the route can be construed. Twopolicies are developed to establish the new link between the new slave and master. The firstone is that master parks an idle active slave based on its traffic record when it constructs linkwith the next forwarding node. As shown in Fig. 12a, M2 parks the idle slave S25 which hasless traffic in the past when it constructs link with A and then can construct link with A. Onthe other hand, the second approach is to park the original bridge node during constructionof links with the next forwarding node and then can establish a new link to new bridge whichwill be the next forwarding node in the route. As shown in Fig. 12b, M2 parks bridge B whenit constructs link with A and then constructs the link with A so that connection of P2 and P1
is maintained.
5.2 Piconet Combination Operation
In route construction phase, since link establishment of the routing path is based on thepage/page scan mechanism, it forms new piconets. For example, as shown in Fig. 13a, thereexists two piconets P1 and P2 in the original scatternet and slave S12 which connects tomaster M1, and constructs link with master M2 to form the new piconet. Then, S12 and M2
become S/M bridges. In order to solve this problem, we use piconet combination operationto combine two piconets to single one. If a slave or an S/S bridge constructs a link with amaster or an S/M bridge and form a new piconet, the master or the S/M bridge executespiconet combination operation to eliminate the newly formed piconet when finishing thelink construction with a slave or an S/Sbridge. Thus, the newly formed piconet can be com-bined to the piconet which has existed. Moreover, the slave or the S/S bridge and the masteror the S/M bridge serves as the slave and a master in the combined piconet, respectively.As shown in Fig. 13b, M2 executes the piconet combination operations to combine its andnewly formed piconets after finishing the link construction with slave S12. Therefore, thenewly formed piconet is eliminated. Moreover, S12 and M2 become the S/S bridge and themaster, respectively.
123
A Location Aware Mobility based Routing Protocol
A
S23
S21S24
SS11
M1
BM2
S22
C
D P2
Master
S/S Bridge
Slave
S25
Parked Node
S23
S21A
S24
SS11
M1
BM2
S22
C
D P2
S25
Master
S/S Bridge
Slave
Parked Node
P1
(a) (b)
Fig. 12 Example of park operation. (a) M2 parks the idle slave S25 which has less traffic flow in the past andthen can connect to A. (b) M2 parks bridge B and then constructs the link to A so that connection of P2 andP1 is maintained
S11S M2
M1
P1P2
D
Master Slave S/M Bridge
S11S M2
M1
P1P2
D
Master Slave S/S Bridge
(b)(a)
Fig. 13 Example of the piconet combination operation. (a) Slave S12 constructs the link with master M2 toform the new piconet. (b) M2 executes the piconet combination operation to combine piconets
M1 M2 M3S B DA
Master
(a) (b)
S/S BridgeSlave
M1
M2
M3A
BD
S
Master
Slave
S/S Bridge
Fig. 14 The basic idea of route optimization. (a) The original topology of the route. (b) The new topologyof the route
5.3 Route Optimization Operation
Let us assume that the routing path is the same as shown in Fig. 14a even after the mobility ofnodes. Then, as shown in Fig. 14b, as node M3 moves into M1’s communication range, theroute length can be further reduced. However, since node M1 and M3 have different hoppingsequence, M1 cannot find M3’s existence.
In order to optimize the route, the basic idea is to piggyback the new location of eachforwarding node in the data packet. Each node checks the locations of all forwarding nodesand applies path reduction operation, if it finds a new forwarding node in its communicationrange. However, there may exist more than one node executing the path reduction operationand hence causes the independent path problem. Therefore, when nodes in the route receivethe new location of other node, they check its location and determine if it can reduce therouting path and record the shorter path information in data packet. After receiving datapacket, the destination node notices the nodes in the shorter routing path to construct theconnection.
123
S.-W. Chang et al.
B
M1
M2
M3
A
D
S
Page scan
Page
Master S/S BridgeSlave
Notice B
M1
M2
M3
A
D
S
Master S/S BridgeSlave
(b)(a)
Fig. 15 The example of route optimization. (a) Node D finds there exist the shorter routing path and notifiesM3and M1 to enter page scan and page states, respectively. (b) M1 constructs connection with M3 and hencethe routing path can be optimized
Since nodes in the route move slowly, they will not create any weak link and the routeis broken. Moreover, each forwarding node periodically puts its location in the data packetso that the data packet always does not store extra information. For example, as shown inFig. 15a, node M3 finds M1 in its communication range and routing path can be reducedto S → M1 → M3 → D. Hence, it puts the path reduction information of M1, M3 in thedata packet. Upon receiving the data packet, node D finds that there exist the shorter routingpath and notices M3 and M1 to enter to page scan and page states, respectively. Finally, asshown in Fig. 15b, M1 constructs the connection with M3 and hence the routing path can beoptimized.
6 Performance Evaluation
In this section we rigorously analyze the performance of our mobility based routing proto-col and compares our location aware mobility based routing protocol (LAMP) with somestandard Bluetooth routing protocols such as RVM [25], LORP [3] and BMP [26].
6.1 Simulation Setups
In our work, we use C++ programming to simulate our protocol. The parameters used in oursimulation are listed in Table 2. In our simulation, initially a connected scatternet with fixednumbers of 100 Bluetooth nodes are taken, which are randomly distributed over a squaredarea of 50 m×50 m and 50 pairs of source and destination nodes are randomly selected toconstruct the route using RVM, LORP, BMR and LAMP. The Constant Bit Rate (CBR)model is used to generate the traffic load for each route and the traffic arrival rate is keptat 100 Kbps. The energy consumption for transmitting or receiving one bit of data is set by0.0763×10−6J. In RVM, a new routing path is searched when the current route is broken.On the other hand, new shortest and backup paths in LAMP and LORP are searched, if thebackup path is broken. The control packets are sent from one node to another and all possiblesuccessful paths between the source and the destination are simulated taking mobility intoconsideration. Thus, the average routing path length is estimated for different numbers ofmobile nodes. In BMR, which is a table driven routing protocol, the master of the sourceknows to which piconet the destination belongs and finds the shortest path to destination.If the scatternet is changed due to nodes mobility, the up-to-date information is notified toeach master. Thus, the master of the source can select the new shortest path when the currentshortest path is broken.
123
A Location Aware Mobility based Routing Protocol
Table 2 Simulation parameters Parameter Value
The number of nodes 100Network size 50×50 m2
Communication range 10 mPairs of source and destination 50Traffic model CBRTraffic arrival rate 100 kbpsEnergy consumption 0.0763×10−6J/bitMobility model Random waypoint model
6.2 Simulation Result
It is to be noted that the proposed LAMP considers the mobility of nodes to construct thebackup path and optimizes the route. Therefore, performance of LAMP is examined in termsof average number of hop counts, average number of control packets, total bandwidth con-sumption ratio and total energy consumption ratio based on the number of mobile nodes andaverage mobility speed. The simulation results are given in detail, as follows.
6.2.1 Average Hop Counts
As shown in Fig. 16, the average hop counts for the different number of mobile nodes aresimulated with different routing protocols that we have considered. The average speed ofeach mobile node is considered as 1.5 m/s in the simulation. From the simulation results, itis observed that the average hop counts of the proposed protocol are less than that of RVMand LORP and similar to BMR. In RVM, LORP and LAMP, new and worse routes are foundafter reestablishing the routes as a result of which average hop counts are raised in theseprotocols, when the number of mobile nodes is increased. The route length of LORP andLAMP is less than that of RVM, since they try to shorten the route length while constructingthe shortest and backup paths. Moreover, LAMP can reduce efficiently the route length byapplying reduction and replacement rules. However, LAMP in some situations cannot con-struct the shortest path in order to construct the disjoint backup path. Therefore, the routelength of LAMP is a little higher than BMR, which can select the new and worse shortestpath.
The average hop counts for the different average mobility speed of the mobile nodes areshown in Fig. 17. All nodes in the scatternet are mobile in the simulation. It is observed thatthe proposed protocol gives tremendous improvement in terms of hop counts for differentaverage mobility speed and is closer to BMR. In RVM, LORP and BMR, they initialize theirprotocols to find new and worse routing paths when search the routing paths. Since, the newshortest path in RVM and BMR and the backup path or the new shortest and backup paths inLORP are longer than the broken route, the average hop counts of all protocol are increasedwhile the average mobile speed is added and the link of the route is broken more rapidly.However, the reduction and replacement rule of LAMP can significantly improve the hopcounts of the shortest and backup paths. Therefore, the route length of LAMP is increasedslightly than BMR, which can often select the shortest route.
6.2.2 Control Packets Overhead
In our simulation, we have analyzed the average number of control packet for different num-ber of mobile nodes as shown in Fig. 18. It is observed that RVM, LORP and LAMP protocols
123
S.-W. Chang et al.
Fig. 16 The average hop counts in different protocols for the different number of mobile nodes
Fig. 17 The average hop counts in different protocols for the different average mobility speed
outperform in terms of the number of control packet as compared to BMR. This is becauseeach master in BMR has to maintain the up-to-date information of the scatternet topology inorder to construct the shortest path. Thus, BMR costs more control packets so that each mas-ter has the same and newest information of the scatternet topology. However, more mobilenodes result that the scattenet structure is easily changed. As a result, BMR requires more andmore control packets when the mobile nodes are increased. On the other hand, RVM, LORPand LAMP also require more number of control packets to reconstruct the route since thehigher number of mobile nodes causes that the route is easily broken. Although the backuppaths constructed by LORP and LAMP can avoid to immediately reconstruct the routes andhence reduce the control packets, RVM does not cost more control packets to reconstruct theroute when broken routes are fewer. Furthermore, LORP and LARP use additional controlpackets to shorten the routing path. As a result, RVM outperforms to LAMP and LORP.Besides, since there are fewer nodes to join to the routes construction in LORP when the
123
A Location Aware Mobility based Routing Protocol
Fig. 18 The average number of control packets in different protocols for the different number of mobile nodes
Fig. 19 The average number of control packets in different protocols for the different average mobility speed
route length is shorter, LORP costs fewer control packets to construct the routes and is lowerthan LARP.
Figure 19 investigates the average number of control packets by varying average mobilityspeed. It is founded that RVM, LORP and LAMP protocols outperform BMR and the controltraffic of all protocols is raised when average mobility speed is increased. Since, the higheraverage speed of the mobile nodes results that scatternet topology is changed frequently;BMR creates large number of control packets to maintain the information of the scatternettopology than RVM, LORP and LAMP. Moreover, since the higher average speed of themobile nodes also causes large number of broken links, RVM requires creating more controlpackets to reconstruct the routes than LORP and LARP, which have constructed the backuppaths. Furthermore, LORP is higher than LAMP when the average mobility speed is largerthan 2 m/s. This is because there are more nodes to join the routes construction in LORP
123
S.-W. Chang et al.
Fig. 20 The total bandwidth consumption ratio in different protocols for the different number of mobile nodes
when the route length becomes longer such that LORP creates more control packets thanLAMP.
6.2.3 Total Bandwidth Consumption Ratio
In Fig. 20, we have compared the total bandwidth consumption ratio in different routing pro-tocols for the different number of mobile nodes. The total bandwidth consumption containsthe used bandwidth of control and data packets. It is observed that LAMP outperforms BMR,RVM and LORP. Since, more control packets has the higher bandwidth consumption and lesshop counts has the lower bandwidth consumption, LAMP and LORP which create similarcontrol packets as RVM and have shorter the route paths can efficiently reduce total band-width consumption. Moreover, LAMP is lower than LORP, since its route length is shorterthan the one of LORP. Since, RVM has highest number of hop counts, its total bandwidthconsumption cannot be efficiently reduced and is higher than LAMP and LORP. On the otherhand, BMR has most total bandwidth consumption due to large number of control packets.Moreover, its total bandwidth consumption is raised when the number of mobile node isincreased such that the ratios of LAMP, LORP and RVM are decreased.
From Fig. 21, it is observed that the total bandwidth consumption of the proposed proto-col is less than that of the RVM, LORP and BMR for the different average mobility speed.This is because LAMP has quite shorter route length and less control overhead no matterthe mobility speed is increased. Besides, the total bandwidth consumption ratio of LAMPand LORP are significantly improved when the average mobility speed is increased. This isbecause BMR has large number of control packets and LORP and LAMP have shorter routelength.
6.2.4 Total Energy Consumption Ratio
Figure 22 measures the total energy consumption ratio in different protocols for differentnumber of mobile nodes. The total energy consumption contains the energy consumption of
123
A Location Aware Mobility based Routing Protocol
Fig. 21 The total bandwidth consumption ratio in different protocols for the different average mobility speed
Fig. 22 The total energy consumption ratio in different protocols for the different number of mobile nodes
control and data packets. It is found that the total energy consumption of LORP is less thanthat of the RVM, LORP and LARP when the number of mobile nodes is less 40, since LORPhas less number of control packets and hop counts and control packets which contain theFFN and FBN in LAMP consume more energy. However, the total energy consumption ofLAMP is least when the number of mobile nodes is equal or larger than 40, since the numberof control packets is similar between LAMP and LORP and LAMP has shorter routes thanLORP. Although, RVM uses least number of control packets, it has higher number of hopcounts and therefore its total energy consumption is slightly improved and is higher thanLAMP and LORP. Besides, BMR has the highest total energy consumption since it requireslarge number of control packets. Moreover, the number of control packet in BMR heavilyincreases with the number of mobile nodes. Therefore, the total energy consumption ratiosof LAMP, LORP and RVM reduce with the number of mobile nodes.
123
S.-W. Chang et al.
Fig. 23 The total energy consumption ratio in different protocols for the different average mobility speed
Figure 23 depicts the total energy consumption ratio in different protocols for the differ-ent average mobility speed. LORP has smallest ratio when the average mobile speed is less1 m/s, since it consumes less energy for control packets than that of LAMP and has shorterrouting path length. However, it is found that the total energy consumption of LAMP obtainssmallest ratio if the average mobility speed is equal or larger than 1 m/s. This is becauseLAMP has quite shorter route length and less control overhead no matter the mobility speedis increased. On the other hand, since BMR has large number of control packets and LORPand LAMP have shorter route length, the total energy consumption ratio of LAMP and LORPsignificantly reduce with the average mobility speed.
7 Conclusions
In this paper, we propose a location aware mobility based routing protocol for an ad hocBluetooth network. We consider location information of the nodes to minimize the numberof hop between the source and the destination. Besides, we propose algorithm how to con-struct the backup paths and to maintain the shortest routing path due to mobility of nodes. Wealso analyze the network bottleneck problems during the construction of route and proposerole switch operation to mitigate these problems. From the simulation result we find thatour protocol outperforms in terms of energy and bandwidth consumption to RVM, LORPand BMR. Since, our protocol supports mobility to construct routing path, it can used in dif-ferent mobility based applications in shopping malls, supermarkets and mobile e-commercescenarios.
References
1. The Bluetooth Specification, http://www.bluetooth.org.2. http://www.ieee802.org/15/3. Chang, C. Y., Yu, G. J., Lin, C.-F., & Wu, T.-T. (2002). Relay reduction and route construction for
Scatternet over Bluetooth radio systems. In Proc. IEEE 16th International Conference on InformationNetworking, II, 2.1–2.10.
4. Perkins, C., & Bhagwat, P. (1994). Destination sequenced distance vector routing for mobile computers.In Proc. ACM SIGCOMM, pp. 234–244, Sep. 1994.
5. Perkins, C., & Royer, E. (1999). Ad-hoc on-demand distance vector routing. In Proc. IEEE Workshop onMobile Computing Systems and Applications, pp. 90–100, Feb. 1999.
6. Giordano, S., Stojmenovic, I., & Blazevic, L. (2003). Position based routing algorithms for ad hoc net-works: A taxonomy. Ad Hoc Wireless Networking, Kluwer Academic Publishers, 2003.
7. Ali, S. (2004). The feasibility of mobile adhoc routing over Bluetooth. Thesis of MEng Degree in Com-puter Science, University College London.
8. Sun, M., Chang, C. K., & Lai, T. H. (2002). A self-routing topology for Bluetooth Scatternets. In Proc.The International Symposium on Parallel Architectures, Philippines, May, 2002.
9. Choi, C. S., & Choi, C. W. (2002). DSR based Bluetooth Scatternet. In Proc. International Conferenceon Circuits/Systems Computers and Communications, Philippines, May, 2002.
10. Mirza, A. B., & Pollard, J. K. Dynamic Source Routing in Bluetooth Personal Area Networks. http://www.ducati.doc.ntu.ac.uk/uksim/journal/Vol-5/No-1&2/MIRZA.pdf
11. Yang, C.-H., & Ruan, J.-W. (2005). On-demand routing for Bluetooth Scatternets subject to devicemobility. In Proceedings of the 19th International Conference on Advanced Information Networking andApplications, 2005.
12. Jian, Z., Fei, Z., & Bin, W. (2005). Cluster based routing in Bluetooth ad-hoc network. http://www.ieeexplore.ieee.org/iel5/10362/32965/01544162.pdf, 2005
13. Werb, J., & Lanzl, C. (1998). A positioning system for finding things indoors. IEEE Spectrum, 9(35),71–78.
14. Priyantha, N.B., Chakraborty, A., & Balakrishnan, H. (2000). The Cricket location-support system.In Proc. Sixth Annual ACM Intl. Conf. on mobile Computing and Networking, 2000.
15. Texas Instruments TIRIS. http://www.ti.com/tiris/default.htm16. Gonzalez-Castano, F. J., Garcia-Reinoso, J. (2003). Survivable Bluetooth Location Networks. In Proc.
IEEE ICC, 2, pp. 1014–1018.17. Vershney, U., Vetter, R. J., & Kalakota, R. (2000). Mobile commerce: A new frontier. Computer, 10,
pp. 32–38.18. Darling, A. (2001). Waiting for the m-commerce explosion. Telecommunication International, 3,
pp. 34–39.19. Engebretson, J. (2006). Opportunities for growth. Wireless Asia.20. Bridgelall, R. (2003). Enabling mobile commerce through pervasive communications with ubiquitous RF
tags. In Proc. IEEE Wireless Communications and Networking, 3, pp. 2041–2046.21. Sangani, K. (2004). RFID sees all. IEEE Review, 50(4), pp. 22–24.22. Ni, L. M., Liu, Y., Lau, Y.-C., & Patil, A. P. (2003). LANDMARC: Indoor location sensing using active
RFID. In Proc. First IEEE International Conference on Pervasive Computing and Communications, 2003.23. Gonzalez-Castano, F. J., & Garcia-Reinoso, J. (2002). Bluetooth Location Networks. In Proc. IEEE
Globecom, 2002.24. Bluetooth Nokia. http://www.nokia.com/phones/6210/bluetooth.html25. Bhagwat, P., & Segall, A. (1999). A routing vector method (RVM) for routing in Bluetooth Scatternets.
In Proc. IEEE Workshop MOMUC, pp. 375–379, Nov. 1999.26. Song, O., Lim, C., & Choi, C.-H. (2004). Mobility Management in Bluetooth ad hoc networks. In Proc.
of 14th Joint Conference on Communications & Information (JCCI), 2004.
Author Biographies
Sheng-Wen Chang received the B.S. degree in Computer Science andInformation Engineering from Tamkang University, Taiwan, in 2004.Since 2005, he was working toward his Ph.D degree and currently heis a PhD candidate in Department of Computer Science and Informa-tion Engineering at Tamkang University. Mr. Chang is a student mem-ber of the IEEE Computer Society, Communication Society and IEICEsociety. He has won lots of scholarships in Taiwan and participated inmany Bluetooth and Wireless Sensor Networking projects and publishedextensively in the wireless networking area. His research interests areWiMAX, wireless sensor networks, Bluetooth radio networks, wirelessmesh networks and Ad Hoc wireless networks, concerning both theo-retic results and algorithms design.
Prasan Kumar Sahoo received Master of Science in Mathematics fromUtkal University, India. He received his M. Tech. degree in ComputerScience from Indian Institute of Technology (IIT), Kharagpur, India andreceived his Ph.D in Mathematics from Utkal University, India in April,2002. He joined in the Software Research Center, National Central Uni-versity, Taiwan in 2001 and later joined as an Assistant Professor in theDepartment of Information Management, Vanung University, Taiwanin 2003. Currently he is working as an Associate Professor in the samedepartment of Vanung University, since 2007. He was the Program Com-mittee Member of MSEAT’2004, MSEAT’2005, WASA’2006, and IEEEAHUC’2006 and 2008. His research interests include coverage prob-lems, modeling and performance analysis of wireless sensor networkand WiMAX.
Chih-Yung Chang received the Ph.D. degree in Computer Science andInformation Engineering from National Central University, Taiwan, in1995. He joined the faculty of the Department of Computer and Infor-mation Science at Aletheia University, Taiwan, as an Assistant Professorin 1997. He was the Chair of the Department of Computer and Informa-tion Science, Aletheia University, from August 2000 to July 2002. Heis currently Professor of Department of Computer Science and Informa-tion Engineering at Tamkang University, Taiwan. Dr Chang served as anAssociate Guest Editor of Journal of Internet Technology (JIT, 2004),Journal of Mobile Multimedia (JMM, 2005), and a member of Edito-rial Board of Tamsui Oxford Journal of Mathematical Sciences (2001–2005). He was an Area Chair of IEEE AINA’2005, Vice Chair of IEEEWisCom 2005 and EUC 2005, Track Chair (Learning Technology in Edu-cation Track) of IEEE ITRE’2005, Program Co-Chair of MNSA’2005,Workshop Co-Chair of INA’2005, MSEAT’ 2003, MSEAT’2004,Publication Chair of MSEAT’2005, and the Program Committee Mem-
ber of USW’2005, WASN’2005, and the 11th Mobile Computing Workshop. Dr. Chang is a member of theIEEE Computer Society, Communication Society and IEICE society. His current research interests includewireless sensor networks, mobile learning, Bluetooth radio systems, Ad Hoc wireless networks, and mobilecomputing.
